The Statistical Method in Inspecting Materials

THE production and inspection of the materials of war involves thousands of observations. Logical action is generally based on the interpretation of many observations. The success of such action depends upon the accuracy of the observations and the soundness of the interpretation placed upon them. There are countless instances where the observer 'sees but perceives not'; that is, the significance of a group of observations is lost, due either to faulty interpretation or to no interpretation at all being placed on the results. This condition is being partly corrected in the United Kingdom and the United States (1) by the promotion of certain principles by Ordnance officials. The term Quality Control has been adopted to describe a system of studying and controlling industrial products by methods of handling observations. In these pages an attempt is made to outline in a non-technical manner some of the ideas used in the handling of observations. Of course the authenticity of the observations determines the worth of any decisions based on them. No matter how brilliant is the theory built on unreliable sampling, it is usually of little practical value. n some fields it is almost impossible to obtain observations free from bias. Let us suppose that observers of speed in miles per hour were available for judgment: (a) a bystander, (b) a motorcycle policeman, and (c) a motorist accused of speeding. The bystander might be unbiased but probably would be very inaccurate. The motorist is usually biased on the low side. The policeman is the only qualified observer and, if over-zealous, he might be biased on the high side. Conclusions drawn from large numbers of biased observations naturally would be of little value.